# Visual Question Answering (Perception) on CV-Bench, BLINK, and RealWorldQA

Source: [Visual Question Answering (Perception) on CV-Bench, BLINK, and RealWorldQA](<https://www.sota2.com/research/sota/visual-question-answering-perception-on-cv-bench-blink-and-realworldqa>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T05:38:53.550Z

Evaluation summarizing performance across the CV-Bench, BLINK, and RealWorldQA datasets for Visual Question Answering tasks, reporting average performance.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Visual Question Answering (Perception) |
| Dataset (including any stated split) | CV-Bench, BLINK, RealWorldQA |
| Metric | CV-Bench Score |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 8 of 8 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [CV-Bench Score](<https://www.sota2.com/research/sota/visual-question-answering-perception-on-cv-bench-blink-and-realworldqa.md?metric=CV-Bench%20Score>): Higher is better; unit: PERCENT.

- [BLINK Score](<https://www.sota2.com/research/sota/visual-question-answering-perception-on-cv-bench-blink-and-realworldqa.md?metric=BLINK%20Score>): Higher is better; unit: PERCENT.

- [RealWorldQA Score](<https://www.sota2.com/research/sota/visual-question-answering-perception-on-cv-bench-blink-and-realworldqa.md?metric=RealWorldQA%20Score>): Higher is better; unit: PERCENT.

- [Average VQA Score](<https://www.sota2.com/research/sota/visual-question-answering-perception-on-cv-bench-blink-and-realworldqa.md?metric=Average%20VQA%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | CV-Bench Score | Result date | Report paper |
| --- | --- | --- | --- |
| LATTE | 60.2 | Not recorded | 68fa7b451e209131cea876c2 |
| LATTE | 57.2 | Not recorded | 68fa7b451e209131cea876c2 |
| LATTE | 56.9 | Not recorded | 68fa7b451e209131cea876c2 |
| Vanilla IT | 56.8 | Not recorded | 68fa7b451e209131cea876c2 |
| Vanilla IT | 52.6 | Not recorded | 68fa7b451e209131cea876c2 |
| Vanilla IT | 52.3 | Not recorded | 68fa7b451e209131cea876c2 |
| LATTE | 51.7 | Not recorded | 68fa7b451e209131cea876c2 |
| Vanilla IT | 50.6 | Not recorded | 68fa7b451e209131cea876c2 |

### LATTE

CV-Bench Score: 60.2. Result ID: 698b83e17a967eddb76f3b3e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7b451e209131cea876c2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7b451e209131cea876c2.md>)

[Implementation](<https://github.com/SalesforceAIResearch/LATTE>)

Recorded evaluation configuration: {"Language / Vision":"Qwen2-7B/SigLIP","Starting checkpoint":"LLaVA-OV Stage 1.5"}. This record does not assert independent reproduction.

### LATTE

CV-Bench Score: 57.2. Result ID: 698b83e17a967eddb76f3b3a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7b451e209131cea876c2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7b451e209131cea876c2.md>)

[Implementation](<https://github.com/SalesforceAIResearch/LATTE>)

Recorded evaluation configuration: {"Language / Vision":"LLaMA3-8B/CLIP","Starting checkpoint":"Mantis Pretrained (Configuration 2)"}. This record does not assert independent reproduction.

### LATTE

CV-Bench Score: 56.9. Result ID: 698b83e17a967eddb76f3b38.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7b451e209131cea876c2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7b451e209131cea876c2.md>)

[Implementation](<https://github.com/SalesforceAIResearch/LATTE>)

Recorded evaluation configuration: {"Language / Vision":"LLaMA3-8B/CLIP","Starting checkpoint":"Mantis Pretrained"}. This record does not assert independent reproduction.

### Vanilla IT

CV-Bench Score: 56.8. Result ID: 698b83e17a967eddb76f3b3d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7b451e209131cea876c2.md>)

Recorded evaluation configuration: {"Language / Vision":"Qwen2-7B/SigLIP","Starting checkpoint":"LLaVA-OV Stage 1.5"}. This record does not assert independent reproduction.

### Vanilla IT

CV-Bench Score: 52.6. Result ID: 698b83e17a967eddb76f3b37.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7b451e209131cea876c2.md>)

Recorded evaluation configuration: {"Language / Vision":"LLaMA3-8B/CLIP","Starting checkpoint":"Mantis Pretrained"}. This record does not assert independent reproduction.

### Vanilla IT

CV-Bench Score: 52.3. Result ID: 698b83e17a967eddb76f3b39.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7b451e209131cea876c2.md>)

Recorded evaluation configuration: {"Language / Vision":"LLaMA3-8B/CLIP","Starting checkpoint":"Mantis Pretrained (Configuration 2)"}. This record does not assert independent reproduction.

### LATTE

CV-Bench Score: 51.7. Result ID: 698b83e17a967eddb76f3b3c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7b451e209131cea876c2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7b451e209131cea876c2.md>)

[Implementation](<https://github.com/SalesforceAIResearch/LATTE>)

Recorded evaluation configuration: {"Language / Vision":"LLaMA3-8B/SigLIP","Starting checkpoint":"Mantis Instruct-tuned"}. This record does not assert independent reproduction.

### Vanilla IT

CV-Bench Score: 50.6. Result ID: 698b83e17a967eddb76f3b3b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7b451e209131cea876c2.md>)

Recorded evaluation configuration: {"Language / Vision":"LLaMA3-8B/SigLIP","Starting checkpoint":"Mantis Instruct-tuned"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.